Robots aren’t coming to take your job!
The best business-driven AI is built with a human in the loop.Pexels / Tara Winstead
Don’t panic! Build human-in-the-loop AI to augment, not automate, writes Ville Hulkko.
Finland is the cradle of knowledge work. Our academia, educators, innovators and engineers shine bright at the top of global benchmarks, and our technology is used daily by nearly every single consumer around the globe.
At the same time, AI-driven automation is widely considered to mainly impact low-knowledge, menial tasks – and usually in a negative fashion. This, however, could not be further from the truth due to a number of key features of AI in the real world.
Real-world business-driven AI is not superior to a human in any way. In fact, generally speaking it might initially make such a vast number of mistakes that removing a human from the working loop could lead to dreadful results. Anyone who has ever used a mediocre chatbot has experienced this. Instead, the best business-driven AI is built with a human in the loop.
In practice, this means offloading tasks that the AI is confident in and using a human to validate any outcome the AI isn’t sure about. When an email bot presents its replies to a human customer service agent before sending it out, the end-user quality is retained, but the process becomes much more efficient.
“By helping AI learn from our everyday tasks we can focus our efforts toward more impact-creating activities.”
This logic applies to knowledge-driven tasks regardless of their complexity. It’s easy to brush off customer service as a job anyone could handle; however, the expertise of a medical doctor takes a different tone. A doctor can view an X-ray image and, in a matter of seconds, determine whether or not a patient is suffering from pneumonia. This skill can easily be taught to AI, even though the knowledge required to perform this task far exceeds the knowledge of a layman. By helping AI learn from our everyday tasks, we can focus our efforts toward more impact-creating activities, such as complex diagnostics and doctor-patient interaction.
Furthermore, this knowledge can be used to train human experts as well. While an expert doctor specialises in X-ray-based data extraction, a junior doctor candidate may struggle to reach the same level of insight. Having these junior operatives collaborate with AI pre-trained by a network of hundreds or thousands of medical doctors can accelerate the learning rate and help them get to a higher skill level faster.
For these reasons, business-driven AI shouldn’t be considered a job-stealing robot. Rather, human-in-the-loop AI is a collaborative digital knowledge co-worker who helps you, learns from you and handles repetitive tasks for you, so you can focus on what truly creates an impact in your job.
This column was first published in November 2018
Photo by Senja Larsen